Supplement to Ethernet
Ethernet is growing fast in industrial process control. This puts many 5-V based fieldbus systems like CAN and Profibus under pressure. Because they need a power supply to convert the supply voltage to 5 V, a timing crystal and some microprocessor assistance to handle the communication stack, they are too expensive and may also be too big to be included in the smallest and most price sensitive devices like push buttons, lamps, micro switches, motor contactors, window openers in cars etc. It is therefore only possible to use these bus systems for distributed I/O where the last meters to the actuators and sensors are either a direct connection or a very simple and cheap master-slave fieldbus like IO-Link, LIN or CAN FD Light. However, Ethernet can do that too at very much the same price, and there may even be Ethernet in the building already, which may save some money for cabling. There are however also a number of limitations with Ethernet compared to simpler systems like Max-i:
- Ethernet too requires a power supply and a timing crystal and the necessary microprocessor must be much more powerfull than for CAN. Max-i gets it supply voltage directly from the bus, use an RC oscillator, which is much more reliable and insensitive to vibrations than a crystal oscillator and doesn't need any microprocessor assistance for the majority of applications.
- Ethernet is not deterministic, so it is necessary with an added layer like Ethernet PowerLink, EtherCAD, ProfiNet, Ethernet/IP, Time Sensitive Networking (TSN) or Sercos. Max-i is fully deterministic and unlike CAN, a high priority device cannot saturate the bus. If all devices want to transmit, they just get through one by one.
- Ethernet requires a TCP/IP stack, which can be large and burdensome to small devices and excludes the use of very small microprocessors. Max-i always handles the entire communication stack in hardware so there is no MAC for the user to warry about.
- The efficiency of Ethernet is extremely low for short messages like the ones typical transmitted in a process plant. The data length is always at least 46 bytes and 26 bytes are needed for preamble, addressing and frame check sequence. Besides, the inter frame gap must be at least 9.6 μs, which corresponds to 12 bytes at 10 Mbps and 120 bytes at 100 Mbps! This reduces the effektive transmission speed considerably compared to simpler bus systems. Max-i is for example able to transmit a 4-bit boolean process value with the long 31-bit identifier in only 7 bytes including a 20-bit CRC-check and a 7-bit Hamming code on the identifier. This is 12 times more efficient than a 10 Mbps Ethernet, which in effect has a communication speed of only 833 kbps in this case, but is still as critical as a 10 Mbps communication! Max-i is also able to transfer a 26-bit analog fixed point value with the long identifier in only 9 bytes, which is over 9 times more efficient than Ethernet. Ethernet may be fast for big data, but it is unnecessary critical for short process values!
- Ethernet uses device addressing. All devices must be assigned a unique physical 48-bit Media Access Control (MAC) address by the manufacturer, which is used on the Ethernet. This address typical contains the 24-bit organizationally unique identifier of the manufacturer and a serial number, but it does not contain information about the various messages, which are transmitted, so this information must be contained in the data field to be able to interpret them. Max-i uses the publisher-subscriber model where it is the various data, which has an identifier, not the devices from which they come. This is much more efficient, and if a numbering system like Plant Numbering System, which is described below, is used, no conversion tables are needed to interpret the data. With this model, many devices can utilize the same messages for example for control panels, displays and lanterns for traffic lights. In Max-i, it is even possible to transmit different data to more devices in the same telegram for even higher efficiency for example for stage lamps, traffic lights, robots, cobots and CNC-machines. On a 25-m line, Max-i is for example able to synchronize over 16 servo motor axes with 32-bit precision to an accuracy of 0.1 μs at an industrial state-of-the-art communication cycle of 400 μs.
- Except for 10BASE-T1S, Ethernet is today a point-to-point communication between a field device and a router or a switch, so every sensor or actuator needs 2 Physical Layers (PHY) - one in each end, 2 Media Access Control devices (MAC) and 1 router or switch channel - plus a lot of cabling - especially if the routers or switches are not distributed in the process plant. Max-i just uses a simple multidrop trunk line and therefore only 1 PHY and MAC, which are entirely implemented in a hardware state machine, which cannot go down like a microprocessor.
- The number of transistors in the necessary microprocessors, Ethernet controllers and routers or switches are very high, which reduces the reliability considerably. For every time the number is doubled and/or the temperature is increased 10 °C, the reliability is reduced to the half. The Max-i controller only use approximately 6000 gates where approximately half is used for the very advanced lighting controller, which may not be used. 6000 gates corresponds approximately to the half of what is used in the smallest ARM M0 microprocessor.
- The maximum power, which can be transferred over twisted pair cables (Power-over-Ethernet - PoE) is only 90 W and this even requires 4 pairs, which is a hassle to mount, and the thin wires break very easily. 90 W excludes the use of PoE for driving actuators and lamps. With Max-i, it is possible with up to 1500 W per segment on standard installation cables.
- The cable length for standard Ethernet is limited to 100 m. Up to 12.8 km is possible with Max-i at 9.6 kbps.
- Some versions of Ethernet use three-level PAM3 signaling (+1, 0, -1), which makes it necessary with two trigger levels, which reduces the effektive signal amplitude to the half and makes it very difficult to optimize these levels as they ought to depend on the signal amplitude. Max-i has only two levels and the triggering point is always in the ideal point in the middle no matter the signal attenuation.
- The low communication voltage of Ethernet combined with the short pulses gives a very low symbol energy (voltage multiplied by current multiplied by time), which reduces the signal/noise ratio considerably. With a 10BASE-T1x peak-to-peak signal amplitude of 2.4 V, a 1.2 V, ±1-to-0 PAM3 transition at a 7.5 MHz clock on a 100 Ω line has a pulse energy of only 4.0 nJ (0.029 W x 0.1333 µs). At 2.667 Mbaud, which is possible on a 25-m line with Max-i, the energy in each pulse is up to 1.7 µJ (6.8 W x 0.25 µs) on a typical 50 Ω, 4-wire, balanced transmission line, which is used for process control.
- Although every part of the Ethernet protocol, routers and switches are standardized and may be familiar to some programmers, it is extremely complex with specifications of several thousand pages and hundreds of parameters to set, and it is very difficult to overview and debug for example for a process electriciant. Max-i is so simple that everybody can learn the protocol in a few hours.
The problem with the limited length of Ethernet and the use of at least two pairs are solved in the new 10 Mbps Single Pair Ethernet (SPE) standards (IEEE 802.3cg-2019) 10BASE-T1L (10 = 10 Mbps, L = long) for communication up to 1-2 km and 10BASE-T1S (S = Short) for multidrop communication between up to 8 devices at maximum 25 m. With these standards, only a single pair is necessary, but the power is limited to 50-60 W. These standards also enables operation at very low power, which makes them usable for Ex areas. However, in practical process control there are typical only 1.5 times more inputs than outputs, so ultra low power on the sensor side does not make much sense if several amperes are needed to drive the actuators. Higher power than approximately 500 mW can easily be handled by means of the High Power Trunk Concept (HPTC), which is used by Max-i. The only requirement is that everything must be sealed and never disconnected during operation, which is anyway the principle used for motors, which may require power in the kW range and use over 400 V. Working on the trunk requires a hot work permit, which is the tradeoff for increased power levels.
10BASE-T1L allows communication up to at least 1 km - typical 1.5-1.7 km, but the use of echo cancelation to allow two-way full duplex communication on a single pair makes the communication even more critical and very complex to start up.
10BASE-T1S is a multidrop line, which enables up to 8 devices to be parallel connected on a single pair, which can be up to 25 m long, but for practical process control where short process values are transmitted, Max-i outperforms 10BASE-T1S on all parameters:
- Max-i is approximately 1.5 times faster and even more if there are more than two devices on the line! The time it takes 10BASE-T1S (and 10BASE-T1L) to transmit a typical process values with an average length of 8 bytes (32-bit identifier and a 50 % mix between 4-bit boolean values and 26-bit analog fixed-point values) is 76.8 µs due to the minimum number of bytes and the long inter frame gap. On a 25-m line, where the Max-i speed is 2.667 Mbaud, each value in average takes 50.8 µs to transmit including inter frame gap.
- If the possibility to send data to more devices in the same message is utilized, the speed differense is even bigger. As written previousy, Max-i is under these circumstances able to synchronize over 16 servo motor axes with 32-bit precision to an accuracy of 0.1 μs at an industrial state-of-the-art communication cycle of 400 µs.
- The response time is approximately 5-10 times shorter! To handle the bus arbitration, 10BASE-T1S uses collision avaidance, which is a kind of token passing where each device gets access one by one in a round robin structure. This gives a response time, which is usually below 500 µs; but it makes the communication even slower compared to Max-i as time is wasted on devices, which have no data to transmit. Max-i is event driven and uses bit-wise bus arbitration so for all practical process control applications where all devices don't need to transmit all the time, the response time is usually close to zero plus the time it takes to transmit the message itself.
- The signal/noise ratio is up to 400 times better as described above (pulse energy
- 4.0 nJ versus1.7 µJ).
- It is posible to connect many more devices. 10BASE-T1S is limited to 8. In practice, Max-i is only limited to the number of devices, it is physical possible to connect.
- It is much simpler.
- The reliability is much higher due to the much simplertechniqueand the lack of crystals.
- Max-i may easily be implemented in even old IC technology, which may be written off. The typical internal clock frequency for 10Base-T1S is in the order of 600 MHz, which requires a fairly advanced and expensive IC technology. Max-i uses only 32 MHz and the present FPGA implementation is made in 130 nm technology.
In the future, there will therefore be a need for two types of bus systems in the industry - Ethernet for high speed data and big data with an added layer to make it deterministic, and an ultra-low-cost, but still high performance system for connection to individual actuators, sensors and lamps. This bus system must:
- Be a multiple master bus so that it is not necessary with an extra layer (distributed device with protocol conversion) between fieldbus and actuators and sensors.
- Be deterministic.
- Be able to run on long lines without being critical.
- Be cheap and small enough for even the smallest and most price sensitive devices as these devices by far make up the majority of devices in most plants. A vendor of a complex device may find it easy to add for example a CANOpen or Profibus Interface if the device already contains a power supply and a sufficient powerful microprocessor with bus controller, and the necessary software stack is available, but selecting a bus system, which is too complex for simple devices on the same bus, will not lead to an optimized total solution.
- Be powerful enough for even very complex and demanding devices like mass flow gauges, motion control, robots and cobots.
- Be able to supply enough power for even big actuators where more amperes may be needed and do it in a cost efficient way. This demand excludes all fieldbus systems, which use special communication cables. A lot of fieldbus systems are really only sensor busses due to the low power or the use of communication methods, which do not allow switching of heavy loads, but a bus, which does not take the actuator side seriously, will not lead to an optimized solution.
With its outstanding versatility, Max-i is so far the only fieldbus, which fulfills all these demands! With Max-i, it is even possible and very easy to utilize the gamma correction and 3rd order smoothing filters of the advanced LED controller to limit the inrush current of for example filament lamps, heaters and DC-motors.
In some ways, Max-i is similar to a PoE solution. They are just designed for two completely different purposes and are therefore very good supplements to each other!
Plant Numbering System (PNS)
In a process plant, it is very important to be able to interpret all process values without corresponding cross-reference tables or else devices from different manufacturers are not able to communicate and it is not possible to do service if you stand in the middle of nowhere and receive an error message from the plant on your mobile phone. This problem may be handled by means of for example OPC UA, which has become quite popular, but it is an extremely complicated standard, which consists of 14 documents with a total of approximately 1250 pages, which requires months to study and a microprocessor to implement and therefore reduces the reliability considerably compared to a pure hardware solution. Instead, Max-i are based on well specified data types, which may be scaled to SI-units, and a new numbering system (PNS) based on alternating characters and digits. PNS utilizes the 31-bit identifier of Max-i to its maximum, but does not require further bytes, which keeps the efficiency very high compared to other fieldbus systems. The idea is:
- To create a numbering system where the machine numbers themselves contain enough information to enable communication with the various devices by means of fieldbus systems and especially Max-i. It shall not be necessary with separate fieldbus addresses, cross-reference tables etc.
- To be able to number the various measurement types like temperature, pressure, level, flow etc. as properties of the equipment to with they belong like tank level and temperature, pump pressure, pipe flow etc. instead of using gauge names. This fits with the publisher-subscriber model of Max-i, XML (Extensible Markup Language) and MQTT and makes it much easier to overlook the process control system. For example, production line PP2 may contain a tank - T9 (PP2T9) - with two level switches, which controls pump PP2P1. In a traditional instrumentation drawing, the two level switches will be named as level switches like for example PP2LG1 (level gauge 1) and PP2LG2, but in this way, it is only possible to tell that two level switches are controlling a pump - not where these level switches are located. However, with PNS it is also possible to use function codes like for example PP2T9L1 and PP2T9L2. Now it is obvious that two levels - L1 and L2 - of tank PP2T9 are controlling the pump. In the same way, the various gauges on a pipe or a tank may also be named as properties of the pipe or the tank like for example PP2WL3F2 (water pipe 3 flow 2) and PP2WL3T1 (water pipe 3 temperature 1). This way of numbering fits perfectly with the hierarchical topic name of MQTT like "PP2/WL3/T1", which may be extended even further to include for example the plant name and the production line like "PlantX/Line1/PP2/WL3/T1". The actual value is then send in the payload field of the PUBLISH message like "50.7".
- To be able to number all parts of the plant including spare parts with the same numbering system. It shall not be necessary with separate numbering systems for machinery, instrumentation, pipes etc.
- To be able to exchange data with all other systems even all over the world where MQTT may be used as communication layer.
- To make it possible to get a record of all components for example for flow gauge A1FG2 just by searching the documentation for this name (without the function code), so that it is not necessary to specify the use of a gauge. In for example numbering systems like DEP, which uses the ISO 3511 standard, the use of a gauge is specified by means of letters following for example F for flow, P for pressure, L for level, T for temperature and Q for any quality parameter like pH, density, power etc. (not specified). For example, the ISO 3511 number #FITBRQCSZA# means a flow indicator (I) and transmitter (T) with a status display (B), a recorder (R) and a totalizer (Q), which is used for control (C), switching (S), trip initiation (Z) and for an alarm (A). It is obvious that this number requires quite a number of bits and a great symbol on the drawing, and if just a small thing is changed like removing the alarm, all documentation, which contains this part, must be rewritten and redrawn. With the ISO 3511 standard, it is hopeless (with a normal search program) to search the documentation for equipment and components if the use is not known because there are so many possibilities.
Note that PNS is only a first draft. Everybody is very welcome to send us any comments and suggestions. The specification is primary based on experiences from feed mills and heating and power stations so there are without doubt many equipment types missing for other types of plants.
The specification of PNS may be downloaded here: www.max-i.org/plantnumbering.pdf
Max-i has all the features needed to make the commissioning as fast and easy as possible.
- By means of the plant numbering system, no cross reference tables are needed.
- Since it is a multi-master bus, which uses bit-wise bus arbitration, any number of debuggers may be added and removed at any time without the need to reconfigure the network.
- A hardware-based acceptance filter makes it easy to select the wanted group of messages and by means of the local/global signal grouping, it is even possible to take single samples from for example a control loop with hundreds of telegrams per second without overloading the debugger or its display.
- It is easy to force all inputs - boolean or analog - to a wanted value. This makes it for example possible to simulate the presence of material and in this way test the plant without any consequence in case of a failure. It also makes it possible to test the response to an exceedance of an alarm limit without the need to for example heat a boiler above its maximum temperature or increase a pressure beyond the limit.
This page is created with WebSite X5 and updated March 27th 2023.